Widar 3.0 is a large-scale public benchmark dataset for WiFi-based gesture recognition that captures Channel State Information (CSI) across diverse users, locations, orientations, and environments, making it one of the most widely adopted datasets for evaluating cross-domain generalizability in WiFi sensing research. It matters because its systematic variation of key environmental and human factors provides a rigorous testbed for assessing how well sensing models transfer across different real-world conditions, directly supporting research into domain adaptation and generalization. The dataset includes multiple subsets corresponding to different body-coordinate gesture representations and is commonly used alongside its predecessors Widar 1.0 and Widar 2.0, with Widar 3.0 being the most comprehensive variant due to its scale and diversity of collection scenarios.
Source Papers
- A Survey on Wi-Fi Sensing Generalizability: Taxonomy, Techniques, Datasets, and Future Research Prospects ↗ — A Survey on Wi-Fi Sensing Generalizability: Taxonomy, Techni
- Deep Learning-Enhanced Human Sensing with Channel State Information: A Survey ↗ — Deep Learning-Enhanced Human Sensing with Channel State Info
- Exposing the CSI: A Systematic Investigation of CSI-based Wi-Fi Sensing Capabilities and Limitations ↗ — Exposing the CSI: A Systematic Investigation of CSI-based Wi
- WiGNN: WiFi-Based Cross-Domain Gesture Recognition Inspired by Dynamic Topology Structure ↗ — WiGNN: WiFi-Based Cross-Domain Gesture Recognition Inspired
- WiMANS: A Benchmark Dataset for WiFi-based Multi-user Activity Sensing ↗ — WiMANS: A Benchmark Dataset for WiFi-based Multi-user Activi